Meta & Children's Online SafetyChat: The leadership decision was limited to NOT building an u13 classifier
Thread View
████ name redacted. Different bars may be the same person or different people.
quoted in an excerpt in the archive.
Sunday 27 September 2020
09:57Antigone Deborah DavisQuick question, on the request for us to improve our age models, what was the decision leadership took on this when they decided to move age verification efforts to device side?
13:29████If I recall correctly, there was no explicit decision on improving our existing age models. The leadership decision was limited to NOT building an u13 classifier and instead focusing our efforts on device side age verification. Core Dimensions continues to get support from product leadership to continue improving their models and to get them globally operational.
15:21Antigone Deborah DavisBut we think they need more support?
16:52████Yes, confirmed with Core Dimensions PM as I was writing the age assurance sections:
"we need (a) more review resources to establish ground truth data on age in priority markets and (b) more engineering resources to invest in model improvements where accuracy is poor (e.g., India)."
17:01Antigone Deborah DavisThank you!!
17:24████Also on improving IG age modeling -- According to Karina, Adam is not on board with prompting to collect age from existing users (unless there's a clear product need). However, Core Dimensions PM did say IG leadership is on board with improving age prediction models for IG in general: "IG leadership (e.g., Adam, Vishal) is fully engaged with getting better IG models, slightly more from analytics perspective (we send updates every few months). It wouldn't hurt to advocate for more investment though given integrity gaps"
17:26████I think I incorporated all of that in my edits to the state of play doc for you to consider including
18:45Antigone Deborah DavisShared an image